mbrossar/ai-imu-dr

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1006 stars · 262 forks observed · 2026-09-21

What it is AI-extracted, prompt v1, taxonomy v1, 2026-09-07, confidence 0.95

Python/PyTorch implementation of the AI-IMU dead reckoning algorithm from the authors' IEEE/ArXiv paper, combining an invariant Kalman filter with a deep neural network that adapts filter noise covariance parameters. It estimates 3D position, velocity, and orientation of wheeled vehicles from raw IMU data, achieving 1.10% translational error on the KITTI odometry benchmark.

Use cases

• estimate vehicle trajectory from IMU data only
• reproduce AI-IMU dead reckoning results on KITTI odometry
• research inertial odometry and learned noise covariance adaptation
• evaluate Kalman filter based state estimation for intelligent vehicles
• benchmark dead reckoning against LiDAR and stereo vision odometry methods

When to choose

• you need IMU-only localization or dead reckoning for wheeled vehicles
• you want to reproduce or extend the AI-IMU paper's research results
• you are studying Kalman filters combined with learned noise parameters
• you have KITTI raw IMU data or similar inertial datasets

When to avoid

• you need a production-ready or real-time embedded localization system
• you expect support for GPS, LiDAR, or visual odometry fusion
• you need a maintained, actively developed codebase with up-to-date dependencies
• your platform is not a standard Python environment

Facets

library · maturity maintenance

machine-learning deep-learning computer-vision data-science autonomous-vehicles robotics machine-learning python windows imu dead-reckoning state-estimation inertial-odometry kalman-filter kitti localization pytorch algorithms linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mbrossar/ai-imu-dr")